DSpace Repository

Bayesian Network Based Student Affect Modeling Framework for an Intelligent Tutoring System

Show simple item record

dc.contributor.advisor Nitin Afzulpurkar (Chairman) en_US
dc.contributor.author Akhtar Hussain en_US
dc.contributor.other Sumanta Guha (Member) en_US
dc.contributor.other Manukid Parchnichkun (Member) en_US
dc.contributor.other Abdul Rehman Abbasi (Member) en_US
dc.contributor.other Fabrice Meriaudeau (External Examiner) en_US
dc.date.accessioned 2015-01-12T10:37:13Z
dc.date.available 2015-01-12T10:37:13Z
dc.date.issued 2012-05 en_US
dc.identifier.other AIT Diss no.CS-12-01
dc.identifier.uri http://www.cs.ait.ac.th/xmlui/handle/123456789/110
dc.description.abstract Human beings have the ability to detect and understand affective/mental states, and other social signals when interacting with each other. This ability or intelligence is an important aspect of their social life in social relationships. Researchers from multi-disciplinary areas have been trying to incorporate this ability or intelligence in computers through body language to make them affective companions of the users for variety of applications. Gestures either intentional or unintentional are very useful to find the underlying mental or affective state of a person in any social interaction. Researchers developing interactive and intelligent computer interfaces are very much interested in extracting meaningful information from variety of gestures, e.g., self-manipulators that include unintended hand-touch-head (face) movements. Predicting human behavior from physical activity is an active area of research for many Affective Computing applications. However, correctly detecting and classifying human body movements specially unintentional body gestures are a research problem. The main problem is occlusion in unintended hand-touch-head (face) gesture’s classification due to similar skin color and texture because when the hand enters in the face region, it merged with the face so difficult to separate the hand from the face region. However, we propose a solution for separating hand(s) from face in varying lighting conditions by generating local binary patterns using force field features in conjunction with Sobel edge operator called (Sobel-LBP). In this dissertation we performed two experiments one with single context and second with multicontext scenarios using real and synthetic data. In our first experiment we used vision based techniques and Bayesian network model for student mental state prediction from unintentional hand-touch-head (face) movements in classroom context. After successful classification of the gestures in the form of binary codes using vision based techniques, we code these different gestures of more than 100 human subjects, and feed these codes manually in three-layered Bayesian network (BN) to infer the probable mental state with particular gesture. The first layer shows the mental states to gestures relationships and the second layer combine the gestures, and SLBP generated binary codes. The proposed scheme when evaluated on a our data set in single context scenario collected in real classroom situation and found promising results with an accuracy of about 85%. The framework will be utilized for developing intelligent tutoring system. In our second experiment we used same techniques for predicting the student mental state in classroom context with multi-context scenarios using real and synthetic data and obtained 75% average accuracy of the predicted result. This result can be improved by increasing the collection of data sets in different contexts in real time situations for training and testing. The results show that our proposed system can also be used for various other applications of Affective Computing. Our proposed system exhibits that in single and as well as in multi-contexts situations, unintentional body gestures can carry useful information about the mental or affective state of the students that can be used to increase the efficiency of various applications such as an intelligent tutoring system by applying an integrated framework of computer vision techniques and Bayesian network model. en_US
dc.language.iso eng en_US
dc.subject Mental states, Social signals, Hand-touch-head/face, Affective Computing, Bayesian network. en_US
dc.subject.lcsh Others en_US
dc.title Bayesian Network Based Student Affect Modeling Framework for an Intelligent Tutoring System en_US
dc.type Dissertation en_US


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search DSpace


Advanced Search

Browse

My Account